RSNA 2007 

Abstract Archives of the RSNA, 2007


SSK12-08

Correlation Analysis in Multiple Frequency Ranges for Functional Brain Connectivity in Normal Pediatric Subjects

Scientific Papers

Presented on November 28, 2007
Presented as part of SSK12: Neuroradiology/Head and Neck (Brain: Physics and Imaging Techniques)

 Trainee Research Prize - Medical Student

Participants

Quan Zhu, Presenter: Nothing to Disclose
Hongtu Zhu PhD, Abstract Co-Author: Nothing to Disclose
John H. Gilmore MD, Abstract Co-Author: Nothing to Disclose
Kathy P. Wilber, Abstract Co-Author: Nothing to Disclose
Jeffrey Keith Smith MD, PhD, Abstract Co-Author: Nothing to Disclose
Weili Lin PhD, Abstract Co-Author: Nothing to Disclose

PURPOSE

Multilevel structural equation model has been utilized to depict putative brain connectivity among brain activation regions resulted from task-related fMRI. Nevertheless, most of the studies to date have focused on adults with relatively little attention toward pediatric subjects (<2yrs). In this work, we focus on the estimation of frequency-dependent correlation matrices to characterize functional connectivity between sensory, motor and visual cortices (SC, MC&VC) in neonates (2-4wks), 1&2yrs old children.

METHOD AND MATERIALS

Nine 2yrs and 10 1yr and 10 neonates were imaged during sleep without sedation using an EPI sequence (TR=2s) on 3T. 6 ROIs were manually drawn in right/left (R/L) SC, MC&VC, respectively. The maximal overlap discrete wavelet transform (MODWT) was applied to decompose signals in 6 frequency bands: B1 0.25-0.5Hz; B2 .125-.25Hz; B3 .063-.125Hz; B4 .031-.063Hz; B5 .016-.031Hz; and B6 0.008-0.016Hz, respectively. Robust PCA was then used to extract the first principal component (PC) in every frequency range for each ROI. Correlation matrix was calculated based on the PCs in each frequency. Multilevel structural equation model was then used to analyze data for each frequency. In this study, only highly correlated regions (correlation coefficient>.4) were considered.

RESULTS

The findings are followings. 1. No correlation was observed in the first two high frequency ranges (B1 & B2). 2. The correlation between R&L VC is consistent throughout B3-B6 for the 1&2yrs groups but not for neonates. 3. The correlation between R&L MC persists for all ages for B4-B6. 4. R&L SC are highly correlated with each other in B4&B5. 4. SC is correlated with MC in B6 for all ages.

CONCLUSION

Cortical functional connectivity are confirmed in spatially segregated regions. Bilateral SC&MC connectivity is observed for all age groups while bilateral VC connectivity is only observed in the 1&2yrs groups. Moreover, there are no significant connections in high frequencies (B1&B2).

CLINICAL RELEVANCE/APPLICATION

The proposed method can be used to investigate frequency-specific functional connectivity between different brain regions during early normal brain development.

Cite This Abstract

Zhu, Q, Zhu, H, Gilmore, J, Wilber, K, Smith, J, Lin, W, Correlation Analysis in Multiple Frequency Ranges for Functional Brain Connectivity in Normal Pediatric Subjects.  Radiological Society of North America 2007 Scientific Assembly and Annual Meeting, November 25 - November 30, 2007 ,Chicago IL. http://archive.rsna.org/2007/5011815.html